Soniox STT v6 adds agent context and voice echo filtering

Soniox has released Soniox STT v6, a hosted speech-to-text API for voice agents. The service is available through the cloud as stt-rt-v6 for streaming and stt-async-v6 for files, supporting more than 60 languages. Soniox says the new model is significantly more accurate than v5, although it has not published independent v6 measurements or a parameter count.

The main addition lets a system send the agent’s spoken text alongside caller audio. Developers enable it with enable_agent_text_input on stt-rt-v6 and transmit timed agent_text JSON frames through the same WebSocket. The model uses this context to interpret brief replies and identify residual agent audio, while continuing to listen for interruptions. Soniox says this filtering supplements, rather than replaces, acoustic echo cancellation.

The release also targets numbers, currencies, dates, IDs, names, specialized terms, accents, code-switching, meetings and diarization. Streaming costs about $0.12 per hour and async processing $0.10, with diarization and language identification included. On the Pipecat benchmark, the predecessor v5 recorded 1.34% semantic WER, 83.3% perfect transcripts and 260 ms median TTFS; v6 has not been tested. The v5 models are scheduled for removal on December 15, 2026, after which requests will route to v6.